Contributor Guide#


This document assumes some familiarity with contributing to open source scientific Python projects using GitHub pull requests. If this does not describe you, you may first want to see the New Contributor FAQ.

Development Workflow#

  1. If you are a first-time contributor:

    • Go to networkx/networkx and click the “fork” button to create your own copy of the project.

    • Clone the project to your local computer:

      git clone
    • Navigate to the folder networkx and add the upstream repository:

      git remote add upstream
    • Now, you have remote repositories named:

      • upstream, which refers to the networkx repository

      • origin, which refers to your personal fork

    • Next, you need to set up your build environment. Here are instructions for two popular environment managers:

      • venv (pip based)

        # Create a virtualenv named ``networkx-dev`` that lives in the directory of
        # the same name
        python -m venv networkx-dev
        # Activate it
        source networkx-dev/bin/activate
        # Install main development and runtime dependencies of networkx
        pip install -r requirements/default.txt -r requirements/test.txt -r requirements/developer.txt
        # (Optional) Install pygraphviz and pydot packages
        # These packages require that you have your system properly configured
        # and what that involves differs on various systems.
        # pip install -r requirements/extra.txt
        # Build and install networkx from source
        pip install -e .
        # Test your installation
        pytest --pyargs networkx
      • conda (Anaconda or Miniconda)

        # Create a conda environment named ``networkx-dev``
        conda create --name networkx-dev
        # Activate it
        conda activate networkx-dev
        # Install main development and runtime dependencies of networkx
        conda install -c conda-forge --file requirements/default.txt --file requirements/test.txt --file requirements/developer.txt
        # (Optional) Install pygraphviz and pydot packages
        # These packages require that you have your system properly configured
        # and what that involves differs on various systems.
        # conda install -c conda-forge --file requirements/extra.txt
        # Install networkx from source
        pip install -e .
        # Test your installation
        pytest --pyargs networkx
    • Finally, we recommend you install pre-commit which checks that your code matches formatting guidelines:

      pre-commit install
  2. Develop your contribution:

    • Pull the latest changes from upstream:

      git checkout main
      git pull upstream main
    • Create a branch for the feature you want to work on. Since the branch name will appear in the merge message, use a sensible name such as ‘bugfix-for-issue-1480’:

      git checkout -b bugfix-for-issue-1480 main
    • Commit locally as you progress (git add and git commit)

  3. Test your contribution:

    • Run the test suite locally (see Testing for details):

      PYTHONPATH=. pytest networkx
    • Running the tests locally before submitting a pull request helps catch problems early and reduces the load on the continuous integration system.

  4. Ensure your contribution is properly formatted.

    • If you installed pre-commit as recommended in step 1, all necessary linting should run automatically at commit time. If there are any formatting issues, the commit will not be successful and linting suggestions will be applied to the patch automatically. Simply git add and git commit a second time to accept the proposed formatting changes.

    • If the above fails for whatever reason, you can also run the linter over the entire codebase with:

      pre-commit run --all-files
  5. Submit your contribution:

    • Push your changes back to your fork on GitHub:

      git push origin bugfix-for-issue-1480
    • Go to GitHub. The new branch will show up with a green Pull Request button—click it.

    • If you want, post on the mailing list to explain your changes or to ask for review.

  6. Review process:

    • Every Pull Request (PR) update triggers a set of continuous integration services that check that the code is up to standards and passes all our tests. These checks must pass before your PR can be merged. If one of the checks fails, you can find out why by clicking on the “failed” icon (red cross) and inspecting the build and test log.

    • Reviewers (the other developers and interested community members) will write inline and/or general comments on your PR to help you improve its implementation, documentation, and style. Every single developer working on the project has their code reviewed, and we’ve come to see it as friendly conversation from which we all learn and the overall code quality benefits. Therefore, please don’t let the review discourage you from contributing: its only aim is to improve the quality of project, not to criticize (we are, after all, very grateful for the time you’re donating!).

    • To update your PR, make your changes on your local repository and commit. As soon as those changes are pushed up (to the same branch as before) the PR will update automatically.


    If the PR closes an issue, make sure that GitHub knows to automatically close the issue when the PR is merged. For example, if the PR closes issue number 1480, you could use the phrase “Fixes #1480” in the PR description or commit message.

  7. Document deprecations and API changes

    If your change introduces any API modifications including deprecations, please make sure the PR has the type: API label.

    To set up a function for deprecation:

    • Use a deprecation warning to warn users. For example:

      msg = "curly_hair is deprecated and will be removed in v3.0. Use sum() instead."
      warnings.warn(msg, DeprecationWarning)
    • Add a warnings filter to networkx/

          "ignore", category=DeprecationWarning, message=<start of message>
    • Add a reminder to doc/developer/deprecations.rst for the team to remove the deprecated functionality in the future. For example:

      * In ``utils/`` remove ``generate_unique_node`` and related tests.


    To reviewers: make sure the merge message has a brief description of the change(s) and if the PR closes an issue add, for example, “Closes #123” where 123 is the issue number.

Divergence from upstream main#

If GitHub indicates that the branch of your Pull Request can no longer be merged automatically, merge the main branch into yours:

git fetch upstream main
git merge upstream/main

If any conflicts occur, they need to be fixed before continuing. See which files are in conflict using:

git status

Which displays a message like:

Unmerged paths:
  (use "git add <file>..." to mark resolution)

  both modified:   file_with_conflict.txt

Inside the conflicted file, you’ll find sections like these:

<<<<<<< HEAD
The way the text looks in your branch
The way the text looks in the main branch
>>>>>>> main

Choose one version of the text that should be kept, and delete the rest:

The way the text looks in your branch

Now, add the fixed file:

git add file_with_conflict.txt

Once you’ve fixed all merge conflicts, do:

git commit


Advanced Git users may want to rebase instead of merge, but we squash and merge PRs either way.


  • All code should have tests.

  • All code should be documented, to the same standard as NumPy and SciPy.

  • All changes are reviewed. Ask on the mailing list if you get no response to your pull request.

  • Default dependencies are listed in requirements/default.txt and extra (i.e., optional) dependencies are listed in requirements/extra.txt. We don’t often add new default and extra dependencies. If you are considering adding code that has a dependency, you should first consider adding a gallery example. Typically, new proposed dependencies would first be added as extra dependencies. Extra dependencies should be easy to install on all platforms and widely-used. New default dependencies should be easy to install on all platforms, widely-used in the community, and have demonstrated potential for wide-spread use in NetworkX.

  • Use the following import conventions:

    import numpy as np
    import scipy as sp
    import matplotlib as mpl
    import matplotlib.pyplot as plt
    import pandas as pd
    import networkx as nx

    After importing sp for scipy:

    import scipy as sp

    access the relevant scipy subpackages from the top-level sp namespace, e.g.:


    Instead of from scipy.sparse import linalg or import scipy.sparse.linalg as spla.

    For example, many libraries have a linalg subpackage: nx.linalg, np.linalg, sp.linalg, sp.sparse.linalg. The above import pattern makes the origin of any particular instance of linalg explicit.

  • Use the decorator not_implemented_for in networkx/utils/ to designate that a function doesn’t accept ‘directed’, ‘undirected’, ‘multigraph’ or ‘graph’. The first argument of the decorated function should be the graph object to be checked.

    @nx.not_implemented_for("directed", "multigraph")
    def function_not_for_MultiDiGraph(G, others):
        # function not for graphs that are directed *and* multigraph
    def function_only_for_Graph(G, others):
        # function not for directed graphs *or* for multigraphs


networkx has an extensive test suite that ensures correct execution on your system. The test suite has to pass before a pull request can be merged, and tests should be added to cover any modifications to the code base. We make use of the pytest testing framework, with tests located in the various networkx/submodule/tests folders.

To run all tests:

$ PYTHONPATH=. pytest networkx

Or the tests for a specific submodule:

$ PYTHONPATH=. pytest networkx/readwrite

Or tests from a specific file:

$ PYTHONPATH=. pytest networkx/readwrite/tests/

Or a single test within that file:

$ PYTHONPATH=. pytest networkx/readwrite/tests/

Use --doctest-modules to run doctests. For example, run all tests and all doctests using:

$ PYTHONPATH=. pytest --doctest-modules networkx

Tests for a module should ideally cover all code in that module, i.e., statement coverage should be at 100%.

To measure the test coverage, run:

$ PYTHONPATH=. pytest --cov=networkx networkx

This will print a report with one line for each file in networkx, detailing the test coverage:

Name                                             Stmts   Miss Branch BrPart  Cover
networkx/                                33      2      2      1    91%
networkx/algorithms/                    114      0      0      0   100%
networkx/algorithms/approximation/       12      0      0      0   100%
networkx/algorithms/approximation/         42      1     18      1    97%

Adding tests#

If you’re new to testing, see existing test files for examples of things to do. Don’t let the tests keep you from submitting your contribution! If you’re not sure how to do this or are having trouble, submit your pull request anyway. We will help you create the tests and sort out any kind of problem during code review.

Image comparison#

To run image comparisons:

$ PYTHONPATH=. pytest --mpl --pyargs networkx.drawing

The --mpl tells pytest to use pytest-mpl to compare the generated plots with baseline ones stored in networkx/drawing/tests/baseline.

To add a new test, add a test function to networkx/drawing/tests that returns a Matplotlib figure (or any figure object that has a savefig method) and decorate it as follows:

def test_barbell():
    fig = plt.figure()
    barbell = nx.barbell_graph(4, 6)
    # make sure to fix any randomness
    pos = nx.spring_layout(barbell, seed=42)
    nx.draw(barbell, pos=pos)
    return fig

Then create a baseline image to compare against later:

$ pytest -k test_barbell --mpl-generate-path=networkx/drawing/tests/baseline


In order to keep the size of the repository from becoming too large, we prefer to limit the size and number of baseline images we include.

And test:

$ pytest -k test_barbell --mpl


We use Sphinx for generating the API and reference documentation.

Pre-built versions can be found at

for both the stable and the latest (i.e., development) releases.


After installing NetworkX and its dependencies, install the Python packages needed to build the documentation by entering the root directory and executing:

pip install -r requirements/doc.txt

Building the example gallery additionally requires the dependencies listed in requirements/extra.txt and requirements/example.txt:

pip install -r requirements/extra.txt
pip install -r requirements/example.txt

To build the HTML documentation, enter doc/ and execute:

make html

This will generate a build/html subdirectory containing the built documentation. If the dependencies in extra.txt and example.txt are not installed, build the HTML documentation without generating figures by using:

make html-noplot

To build the PDF documentation, enter:

make latexpdf

You will need to have LaTeX installed for this.


sphinx supports many other output formats. Type make without any arguments to see all the built-in options.

Adding examples#

The gallery examples are managed by sphinx-gallery. The source files for the example gallery are .py scripts in examples/ that generate one or more figures. They are executed automatically by sphinx-gallery when the documentation is built. The output is gathered and assembled into the gallery.

Building the example gallery locally requires that the additional dependencies in requirements/example.txt be installed in your development environment.

You can add a new plot by placing a new .py file in one of the directories inside the examples directory of the repository. See the other examples to get an idea for the format.


Gallery examples should start with plot_, e.g.

General guidelines for making a good gallery plot:

  • Examples should highlight a single feature/command.

  • Try to make the example as simple as possible.

  • Data needed by examples should be included in the same directory and the example script.

  • Add comments to explain things are aren’t obvious from reading the code.

  • Describe the feature that you’re showcasing and link to other relevant parts of the documentation.

Adding References#

If you are contributing a new algorithm (or an improvement to a current algorithm), a reference paper or resource should also be provided in the function docstring. For references to published papers, we try to follow the Chicago Citation Style. The quickest way of generating citation in this style is by searching for the paper on Google Scholar and clicking on the cite button. It will pop up the citation of the paper in multiple formats, and copy the Chicago style.

We prefer adding DOI links for URLs. If the DOI link resolves to a paywalled version of the article, we prefer adding a link to the arXiv version (if available) or any other publicly accessible copy of the paper.

An example of a reference:

.. [1] Cheong, Se-Hang, and Yain-Whar Si. "Force-directed algorithms for schematic drawings and
placement: A survey." Information Visualization 19, no. 1 (2020): 65-91.

If the resource is uploaded as a PDF/DOCX/PPT on the web (lecture notes, presentations) it is better to use the wayback machine to create a snapshot of the resource and link the internet archive link. The URL of the resource can change, and it creates unreachable links from the documentation.

Using Math Formulae and Latex Formatting in Documentation#

When working with docstrings that contain math symbols or formulae use raw strings (r""") to ensure proper rendering. While LaTeX formatting can improve the appearance of the rendered documentation, it’s best to keep it simple and readable.

An example of a math formula:

.. math::
    Ax = \lambda x
\[Ax = \lambda x\]

Some inline math:

These are Cheeger's Inequalities for \d-Regular graphs:
$\frac{d- \lambda_2}{2} \leq h(G) \leq \sqrt{2d(d- \lambda_2)}$

These are Cheeger’s Inequalities for d-Regular graphs: \(\frac{d- \lambda_2}{2} \leq h(G) \leq \sqrt{2d(d- \lambda_2)}\)


Please report bugs on GitHub.


All interactions with the project are subject to the NetworkX code of conduct.

We also follow these policies: